spark VectorizedDeltaBinaryPackedReader 源码

spark VectorizedDeltaBinaryPackedReader 代码

文件路径：/sql/core/src/main/java/org/apache/spark/sql/execution/datasources/parquet/VectorizedDeltaBinaryPackedReader.java

/*
 * Licensed to the Apache Software Foundation (ASF) under one or more
 * contributor license agreements.  See the NOTICE file distributed with
 * this work for additional information regarding copyright ownership.
 * The ASF licenses this file to You under the Apache License, Version 2.0
 * (the "License"); you may not use this file except in compliance with
 * the License.  You may obtain a copy of the License at
 *
 *    http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
package org.apache.spark.sql.execution.datasources.parquet;

import java.io.IOException;
import java.math.BigInteger;
import java.nio.ByteBuffer;
import java.util.Arrays;

import org.apache.parquet.Preconditions;
import org.apache.parquet.bytes.ByteBufferInputStream;
import org.apache.parquet.bytes.BytesUtils;
import org.apache.parquet.column.values.bitpacking.BytePackerForLong;
import org.apache.parquet.column.values.bitpacking.Packer;
import org.apache.parquet.io.ParquetDecodingException;
import org.apache.spark.sql.catalyst.util.RebaseDateTime;
import org.apache.spark.sql.execution.datasources.DataSourceUtils;
import org.apache.spark.sql.execution.vectorized.WritableColumnVector;

/**
 * An implementation of the Parquet DELTA_BINARY_PACKED decoder that supports the vectorized
 * interface. DELTA_BINARY_PACKED is a delta encoding for integer and long types that stores values
 * as a delta between consecutive values. Delta values are themselves bit packed. Similar to RLE but
 * is more effective in the case of large variation of values in the encoded column.
 * <p>
 * DELTA_BINARY_PACKED is the default encoding for integer and long columns in Parquet V2.
 * <p>
 * Supported Types: INT32, INT64
 * <p>
 *
 * @see <a href="https://github.com/apache/parquet-format/blob/master/Encodings.md#delta-encoding-delta_binary_packed--5">
 * Parquet format encodings: DELTA_BINARY_PACKED</a>
 */
public class VectorizedDeltaBinaryPackedReader extends VectorizedReaderBase {

  // header data
  private int blockSizeInValues;
  private int miniBlockNumInABlock;
  private int totalValueCount;
  private long firstValue;

  private int miniBlockSizeInValues;

  // values read by the caller
  private int valuesRead = 0;

  // variables to keep state of the current block and miniblock
  private long lastValueRead;  // needed to compute the next value
  private long minDeltaInCurrentBlock; // needed to compute the next value
  // currentMiniBlock keeps track of the mini block within the current block that
  // we read and decoded most recently. Only used as an index into
  // bitWidths array
  private int currentMiniBlock = 0;
  private int[] bitWidths; // bit widths for each miniBlock in the current block
  private int remainingInBlock = 0; // values in current block still to be read
  private int remainingInMiniBlock = 0; // values in current mini block still to be read
  private long[] unpackedValuesBuffer;

  private ByteBufferInputStream in;

  // temporary buffers used by readByte, readShort, readInteger, and readLong
  private byte byteVal;
  private short shortVal;
  private int intVal;
  private long longVal;

  @Override
  public void initFromPage(int valueCount, ByteBufferInputStream in) throws IOException {
    Preconditions.checkArgument(valueCount >= 1,
        "Page must have at least one value, but it has " + valueCount);
    this.in = in;
    // Read the header
    this.blockSizeInValues = BytesUtils.readUnsignedVarInt(in);
    this.miniBlockNumInABlock = BytesUtils.readUnsignedVarInt(in);
    double miniSize = (double) blockSizeInValues / miniBlockNumInABlock;
    Preconditions.checkArgument(miniSize % 8 == 0,
        "miniBlockSize must be multiple of 8, but it's " + miniSize);
    this.miniBlockSizeInValues = (int) miniSize;
    // True value count. May be less than valueCount because of nulls
    this.totalValueCount = BytesUtils.readUnsignedVarInt(in);
    this.bitWidths = new int[miniBlockNumInABlock];
    this.unpackedValuesBuffer = new long[miniBlockSizeInValues];
    // read the first value
    firstValue = BytesUtils.readZigZagVarLong(in);
  }

  // True value count. May be less than valueCount because of nulls
  int getTotalValueCount() {
    return totalValueCount;
  }

  @Override
  public byte readByte() {
    readValues(1, null, 0, (w, r, v) -> byteVal = (byte) v);
    return byteVal;
  }

  @Override
  public short readShort() {
    readValues(1, null, 0, (w, r, v) -> shortVal = (short) v);
    return shortVal;
  }

  @Override
  public int readInteger() {
    readValues(1, null, 0, (w, r, v) -> intVal = (int) v);
    return intVal;
  }

  @Override
  public long readLong() {
    readValues(1, null, 0, (w, r, v) -> longVal = v);
    return longVal;
  }

  @Override
  public void readBytes(int total, WritableColumnVector c, int rowId) {
    readValues(total, c, rowId, (w, r, v) -> w.putByte(r, (byte) v));
  }

  @Override
  public void readShorts(int total, WritableColumnVector c, int rowId) {
    readValues(total, c, rowId, (w, r, v) -> w.putShort(r, (short) v));
  }

  @Override
  public void readIntegers(int total, WritableColumnVector c, int rowId) {
    readValues(total, c, rowId, (w, r, v) -> w.putInt(r, (int) v));
  }

  // Based on VectorizedPlainValuesReader.readIntegersWithRebase
  @Override
  public final void readIntegersWithRebase(
      int total, WritableColumnVector c, int rowId, boolean failIfRebase) {
    readValues(total, c, rowId, (w, r, v) -> {
      if (v < RebaseDateTime.lastSwitchJulianDay()) {
        if (failIfRebase) {
          throw DataSourceUtils.newRebaseExceptionInRead("Parquet");
        } else {
          w.putInt(r, RebaseDateTime.rebaseJulianToGregorianDays((int) v));
        }
      } else {
        w.putInt(r, (int) v);
      }
    });
  }

  @Override
  public void readUnsignedIntegers(int total, WritableColumnVector c, int rowId) {
    readValues(total, c, rowId, (w, r, v) -> {
      w.putLong(r, Integer.toUnsignedLong((int) v));
    });
  }

  @Override
  public void readUnsignedLongs(int total, WritableColumnVector c, int rowId) {
    readValues(total, c, rowId, (w, r, v) -> {
      w.putByteArray(r, new BigInteger(Long.toUnsignedString(v)).toByteArray());
    });
  }

  @Override
  public void readLongs(int total, WritableColumnVector c, int rowId) {
    readValues(total, c, rowId, WritableColumnVector::putLong);
  }

  @Override
  public final void readLongsWithRebase(
      int total, WritableColumnVector c, int rowId, boolean failIfRebase, String timeZone) {
    readValues(total, c, rowId, (w, r, v) -> {
      if (v < RebaseDateTime.lastSwitchJulianTs()) {
        if (failIfRebase) {
          throw DataSourceUtils.newRebaseExceptionInRead("Parquet");
        } else {
          w.putLong(r, RebaseDateTime.rebaseJulianToGregorianMicros(timeZone, v));
        }
      } else {
        w.putLong(r, v);
      }
    });
  }

  @Override
  public void skipBytes(int total) {
    skipValues(total);
  }

  @Override
  public void skipShorts(int total) {
    skipValues(total);
  }

  @Override
  public void skipIntegers(int total) {
    skipValues(total);
  }

  @Override
  public void skipLongs(int total) {
    skipValues(total);
  }

  private void readValues(int total, WritableColumnVector c, int rowId,
      IntegerOutputWriter outputWriter) {
    if (valuesRead + total > totalValueCount) {
      throw new ParquetDecodingException(
          "No more values to read. Total values read:  " + valuesRead + ", total count: "
              + totalValueCount + ", trying to read " + total + " more.");
    }
    int remaining = total;
    // First value
    if (valuesRead == 0) {
      outputWriter.write(c, rowId, firstValue);
      lastValueRead = firstValue;
      rowId++;
      remaining--;
    }
    while (remaining > 0) {
      int n;
      try {
        n = loadMiniBlockToOutput(remaining, c, rowId, outputWriter);
      } catch (IOException e) {
        throw new ParquetDecodingException("Error reading mini block.", e);
      }
      rowId += n;
      remaining -= n;
    }
    valuesRead = total - remaining;
  }


  /**
   * Read from a mini block.  Read at most 'remaining' values into output.
   *
   * @return the number of values read into output
   */
  private int loadMiniBlockToOutput(int remaining, WritableColumnVector c, int rowId,
      IntegerOutputWriter outputWriter) throws IOException {

    // new block; read the block header
    if (remainingInBlock == 0) {
      readBlockHeader();
    }

    // new miniblock, unpack the miniblock
    if (remainingInMiniBlock == 0) {
      unpackMiniBlock();
    }

    // read values from miniblock
    int valuesRead = 0;
    for (int i = miniBlockSizeInValues - remainingInMiniBlock;
        i < miniBlockSizeInValues && valuesRead < remaining; i++) {
      // calculate values from deltas unpacked for current block
      long outValue = lastValueRead + minDeltaInCurrentBlock + unpackedValuesBuffer[i];
      lastValueRead = outValue;
      outputWriter.write(c, rowId + valuesRead, outValue);
      remainingInBlock--;
      remainingInMiniBlock--;
      valuesRead++;
    }

    return valuesRead;
  }

  private void readBlockHeader() {
    try {
      minDeltaInCurrentBlock = BytesUtils.readZigZagVarLong(in);
    } catch (IOException e) {
      throw new ParquetDecodingException("Can not read min delta in current block", e);
    }
    readBitWidthsForMiniBlocks();
    remainingInBlock = blockSizeInValues;
    currentMiniBlock = 0;
    remainingInMiniBlock = 0;
  }

  /**
   * mini block has a size of 8*n, unpack 32 value each time
   *
   * see org.apache.parquet.column.values.delta.DeltaBinaryPackingValuesReader#unpackMiniBlock
   */
  private void unpackMiniBlock() throws IOException {
    Arrays.fill(this.unpackedValuesBuffer, 0);
    BytePackerForLong packer = Packer.LITTLE_ENDIAN.newBytePackerForLong(
        bitWidths[currentMiniBlock]);
    for (int j = 0; j < miniBlockSizeInValues; j += 8) {
      ByteBuffer buffer = in.slice(packer.getBitWidth());
      if (buffer.hasArray()) {
        packer.unpack8Values(buffer.array(),
          buffer.arrayOffset() + buffer.position(), unpackedValuesBuffer, j);
      } else {
        packer.unpack8Values(buffer, buffer.position(), unpackedValuesBuffer, j);
      }
    }
    remainingInMiniBlock = miniBlockSizeInValues;
    currentMiniBlock++;
  }

  // From org.apache.parquet.column.values.delta.DeltaBinaryPackingValuesReader
  private void readBitWidthsForMiniBlocks() {
    for (int i = 0; i < miniBlockNumInABlock; i++) {
      try {
        bitWidths[i] = BytesUtils.readIntLittleEndianOnOneByte(in);
      } catch (IOException e) {
        throw new ParquetDecodingException("Can not decode bitwidth in block header", e);
      }
    }
  }

  private void skipValues(int total) {
    // Read the values but don't write them out (the writer output method is a no-op)
    readValues(total, null, -1, (w, r, v) -> {});
  }

}

相关信息

spark 源码目录

相关文章

spark ParquetColumnVector 源码

spark ParquetDictionary 源码

spark ParquetFooterReader 源码

spark ParquetReadState 源码

spark ParquetVectorUpdater 源码

spark ParquetVectorUpdaterFactory 源码

spark SpecificParquetRecordReaderBase 源码

spark VectorizedColumnReader 源码

spark VectorizedDeltaByteArrayReader 源码

spark VectorizedDeltaLengthByteArrayReader 源码